So now I have to type “news.” before it will fill news.ycombinator.com, because it started showing me the wrong option and I kept inadvertently picking it because it used to give me the right one.

There doesn’t seem to be any mechanism for it to realize “every time this happens he closes that website immediately and goes to a different one that we ranked #2 in the autocomplete search.” So until I retrain my muscle memory I’ll just keep on mistraining the autocomplete even further.

This is a disheartening yet frighteningly common problem with UI automation. It doesn't learn enough to actually be aware of the expected result, yet it can't be programmed to behave exactly as you want or need.

Automation in user interfaces should be exactly the opposite. It should try to learn your repeated actions so that you can perform them automatically, yet if the learning goes wrong, the user should be in control at all times and override the automatic deduced actions.

I use them at least a few times a year, though far less than I used to when I was on the interstate a lot more. Any time traffic rapidly drops more than about 20mph, I light up my hazards to let following traffic know, "Yes, you see brake lights, no, they're not just people scrubbing a few miles an hour off - get on your brakes now!" Probably a habit I picked up from truckers, a lot of them do this for the very understandable reason that a big rig doesn't stop on a dime, and even if they will, you're likely to unstack your cargo in the process.

Super infrequently used compared to other buttons, but also not something I really have the time go sorting through menus for when I need it. And neither do I trust the car's automatic systems to turn them on for me.

Though, if I could dream, we'd use LED brake lights to encode braking intensity somehow. The car knows if I'm barely touching the pedal to cancel cruise control and light up the brakes, or if I've just mashed them to the ABS actuation point, but the brake lights don't encode any of this useful info. You couldn't rely on it for car behavior (trivial to spoof, and get rid of tailgaters), but it would be an additional useful input for driving - "Woah, hey, that car in front of you just nearly locked up their wheels, radar data agrees, slow down!"

I suppose regulations do not allow it in the U.S. but at least here it can be programmed on/off via service terminals.

Random example from YouTube: https://www.youtube.com/watch?v=dJk-rJpgF1o&t=108s

Maybe the concern is that in heavy traffic a sea of flashing lights will overwhelm / distract people too much? Or cause seizures? I'm really not sure.

I don't know the legality of them, but nobody ever complained on the motorcycles.

Most German cars will flash the brake lights when you push the brakes hard, and they will also automatically turn on hazard lights if you brake hard to a (near) full stop (assuming you were going some minimum speed of 50 or 70 km/h).

ABS stepping in is not a requirement as far as I could tell (had this a few times when erring on the side of caution when the light turned yellow). Not requiring ABS makes sense because even if you’re driving on proper roads and your recent German car has great tires and brakes and doesn’t need ABS to decelerate quickly, the truck behind you probably still needs a bit more time and early warning to avoid a collision.

For the type of situation that you mentioned, it’s common to design counter metrics or context dependent metrics. Define the “really important” situation, see how often people can use the necessary button in that time. Define your goals based on that ratio and you are on the right track.

The thing is, interfaces are limited. I think there’s a reasonable discussion around how important different features are, especially when they are rarely used in general.

Again, I’m not too familiar with Tesla, and it sounds like they made choices different than you think they should. But it’s not the approach that is the problem - setting bad goals or bad strategies is at issue.

I suppose a way is to correlate use of the button with activity that may indicate an emergency situation. For example strong accelerations, quick movement, etc... Some touchscreen can also sense how strong the push is, and we have a tendency to push harder on important buttons.

This is stupid. It's a car and you need controls accessible in an instant, not hidden behind a login page and captcha.

Rarely used buttons may be hard to find and yet very important in specific situations. If car companies really want to use some form of AI in the UI, that's going to be bad in almost all cases, except if they use it to help you find that important but rarely used buttons in those rare situations where you need it.

Of course it should also still be immediately accessible in its normal spot.

There's a reason things like planes and industrial equipment don't have dynamic interfaces.